Berkeley-CHAI All Hands

The future of a beautiful online commons depends on us. AI is a phase transition in how technology shapes our lives — it has already swept through the space of text and ideas, and its arrival in the web's graphical spaces is imminent. In the past, bots moved through the web via structured channels; soon, agentic systems will meaningfully act in what were primarily human spaces, becoming all but indistinguishable from us.We made many mistakes we still live with: closed platforms, surveillance by default and fundamental protocols owned by the few. We get to choose differently for AI — we have better political and digital tools. In this founding era of the agentic web, how shall we build the protocols that underlie the web we want to be in? This is an invitation to have a conversation about how the development of technology can shift actor's influence in the AI landscape. We'll briefly introduce a project we've been working on with Berkeley and Mozilla: an open-source environment for robust evaluation of agents on the Web. We'll then turn to an upcoming effort that aims to inspire open-source protocols for web-agent behaviour.